Classification of Carcinoma in Images Using Efficientnet

Early detection of carcinoma has the potential to improve patient care and thereby save the life of a patient. The use of deep neural nets to automatically detect carcinoma in MRI images can speed up and help with early diagnosis. However, the application of deep learning to analyze carcinoma images remains a challenging task due to inherent complexities present in a MRI image such as fine-grained features, varying resolution, image quality, slices, occlusion, and lack of available data to sufficiently train the model. Recent advancements in CNN architecture, such as variants of EfficientNet, offer new opportunities to improve carcinoma detection. We train a compact deep learning model named EfficientNetV2-M to classify four stages of carcinoma in images. Results show our carcinoma classification model achieves an overall high validation accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 7 \%}$</tex>. The validation accuracy of four individual stages lies between 95.7 % and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 9. 1 \%}$</tex>. Our research underscores the delicate balance between optimizing model performance and mitigating overfitting.